Objective
Accurate prediction of whether a nulliparous woman will have a vaginal delivery would be a major advance in obstetrics. The objective of the study was to develop such a model based on maternal characteristics and the results of intrapartum ultrasound.
Study Design
One hundred twenty-two nulliparous women in the first stage of labor were included in a prospective observational 2-centre study. Labor was classified as prolonged according to the respective countries’ national guidelines. Fetal head position was assessed with transabdominal ultrasound and cervical dilatation by digital examination, and transperineal ultrasound was used to determine head-perineum distance and the presence of caput succedaneum. The subjects were divided into a testing set (n = 61) and a validation set (n = 61) and a risk score derived using multivariable logistic regression with vaginal birth as the outcome, which was dichotomized into no/cesarean delivery and yes/vaginal birth. Covariates included head-perineum distance, caput succedaneum, and occiput posterior position, which were dichotomized respectively into the following: ≤40 mm, >40 mm, <10 mm, ≥10 mm, and no, yes. Maternal age, gestational age, and maternal body mass index were included as continuous covariates.
Results
Dichotomized score is significantly associated with vaginal delivery ( P = .03). Women with a score above the median had greater than 10 times the odds of having a vaginal delivery as compared with those with a score below the median. The receiver-operating characteristic curve showed an area under the curve of 0.853 (95% confidence interval, 0.678–1.000).
Conclusion
A risk score based on maternal characteristics and intrapartum findings can predict vaginal delivery in nulliparous women in the first stage of labor.
Prolonged nulliparous labor is responsible for 30% of all cesarean deliveries. Although risk factors such as induction of labor and occiput posterior position increase the likelihood of cesarean delivery, there is no way of predicting this outcome with precision based on clinical examination alone.
Digital vaginal examination is the key method used worldwide to define the diagnosis of labor and its progress, the findings being plotted manually on a graphical representation of the progress of labor: a partogram. The partogram has been developed to aid the recognition of failure to progress in labor and has formed the basis for national guidelines. However, there is much discussion regarding its applicability, resulting in the American Association of Obstetricians and Gynecologists issuing guidelines on labor management in 2014.
Digital vaginal examination is also intrusive, subjective, inconsistent, and associated with infection. Our group has recently described an ultrasound-based partogram (sonopartogram) based on nonintrusive intrapartum ultrasound observations, using transabdominal and transperineal ultrasound.
Transabdominal ultrasound can be used to determine the fetal head position in labor with a higher degree of accuracy than by classical clinical palpation. Transperineal ultrasound, first described in the mid-1990s, is a 2-dimensional, noninvasive, objective, simple method of scanning, placing a transducer outside the vagina. It can be used to determine the angle of progression of the fetal head, head-perineum distance (HPD), and the degree of caput succedaneum.
Increasingly, models including maternal characteristics and ultrasound findings are used in risk prediction with womens’ health, for example in prenatal Down’s syndrome screening, and in preeclampsia risk prediction. Recently fetal Doppler assessment of the cerebroumbilical ratio prior to the onset of labor has been shown to identify those fetuses at high risk of emergency cesarean delivery because of fetal distress. Knowledge of the likelihood of cesarean delivery, especially in prolonged labor, has the potential to inform obstetric care and expectations of a woman in labor. Taking basic maternal characteristics, clinical and ultrasound observations, we aimed to develop a model for prediction of vaginal birth in the nulliparas in the first stage of labor.
Materials and Methods
A prospective observational 2-center study investigating the clinical value of ultrasound in labor was performed at Stavanger University Hospital (Stavanger, Norway) and at Addenbrooke’s Hospital (Cambridge, United Kingdom) in 2012–2013.
Nulliparous women with a live singleton pregnancy and cephalic presentation in the first stage of labor at term (≥37 weeks) were eligible for the study. Clinical results from the study population have been published previously with respect of the individual component measurements. For this study, patients were included in which all measurements were available specifically in this case including caput succedaneum measurements. Ethics Committee approval was obtained in Norway (REK 2011/731) and in the United Kingdom (11/EE/064), and all women gave written consent.
Prolonged first stage of labor was defined according to national guidelines. World Health Organization recommendations were used in Norway when cervical dilatation crossed the action line (4 hours from the alert line). In the United Kingdom, National Institutes of Health and Clinical Excellence guidelines (dilatation of <2 cm in 4 hours) were used.
Gestational age was calculated from a second-trimester scan in Norway and from the first-trimester scan in the United Kingdom. Maternal age was taken as age at the time of labor and booking maternal body mass index (BMI) was entered in the predictive model. Vaginal delivery included spontaneous and assisted operative deliveries using vacuum extraction and/or forceps.
During the ultrasound examination, women were in the supine position with flexed hips and knees, membranes had ruptured and the bladder was empty as previously described. The birth attendant (midwife or doctor) performed a digital vaginal examination assessing cervical dilatation and fetal head descent. The descent was categorized using the World Health Organization classification of fetal head station with the ischial spines as reference point 0, –5 at the pelvic inlet, and +5 at the pelvic outlet.
An obstetrician or midwife performed all the ultrasound examinations using Voluson i (GE, Zipf, Austria) with a 3.5–7.5 MHz three-dimensional curved multifrequency transabdominal transducer in Norway and Samsung Medison Accuvix XG (Samsung Medison, Medical Imaging Systems Ltd, London, UK) equipment with a 4–6 MHz convex transabdominal transducer in the United Kingdom. Birth attendants were blinded to the ultrasound findings and these findings were also not disclosed to the parents.
Fetal position was assessed with a transabdominal scan and recorded with half-hourly divisions. Positions >03.30 and <08.30 were recorded as the occiput posterior position and other positions as the nonocciput position. Thereafter the transducer was placed transperineally at the level of the posterior fourchette in a transverse position, and the HPD was measured as previously described ( Figure 1 ). Caput succedaneum was measured by the skin-skull distance in a transperineal scan with the probe held sagittally ( Figure 2 ).
Statistical analyses
A predictive model was tested using R version 2.15, which included known variables at the time of the ultrasound examination such as maternal age and BMI. First, the study population (n = 122) was split into a testing set (n = 61) and a validation set (n = 61) using blocked randomization to ensure an equal number of prolonged labor cases in both sets. This was to make our model more externally generalizable because our study population had an overrepresentation of women with prolonged labor.
Table 1 shows the demographic and covariate information by study set as well as basic univariates comparing the distribution of patients by set using Pearson’s χ 2 test for categorical variables and a Student 2-sample t test with equal variance for continuous variables. Vaginal birth was dichotomized into no/cesarean delivery and into yes/vaginal birth. Head perineum distance, caput, occiput posterior position, and prolonged labor were all dichotomized into the following, respectively: ≤40 mm; >40 mm; <10 mm; ≥10 mm; no, yes; and no, yes.
| Characteristic | Test Set (n = 61) | Validation Set (n = 61) | P value a |
|---|---|---|---|
| Maternal age, mean (SD) | 29.6 (5.1) | 29.2 (5.3) | .69 |
| Gestational age, mean (SD) | 40.0 (1.3) | 40.3 (0.83) | .16 |
| Maternal BMI, mean (SD) | 24.9 (3.9) | 25.0 (4.2) | .96 |
| Cervical dilation, mean (SD) | 6.3 (1.6) | 6.1 (1.6) | .53 |
| Vaginal birth | |||
| Cesarean delivery (0) | 20 | 9 | .04 b |
| Vaginal birth (1) | 41 | 52 | |
| Head perineum distance ≤40 mm | |||
| ≤40 mm (0) | 26 | 24 | .85 |
| >40 mm (1) | 35 | 37 | |
| Caput 10 mm | |||
| <10 mm (0) | 34 | 29 | .47 |
| ≥10 mm (1) | 27 | 32 | |
| Occiput posterior position | |||
| No (0) | 37 | 42 | .99 |
| Yes (1) | 18 | 19 | |
| Unknown | 6 | 0 | |
| Prolonged labor | |||
| No (0) | 16 | 16 | 1.00 |
| Yes (1) | 45 | 45 |
a Using Pearson’s χ 2 test and Student 2-sample t test with equal variance as appropriate
Maternal age, gestational age, clinically assessed cervical dilation, and maternal BMI were all kept as continuous variables for the prediction model. A risk score was constructed in the testing set using multivariable logistic regression with vaginal birth as the outcome. The covariates included in the model included the head perineum distance, caput, occiput posterior position, prolonged labor, maternal age, gestational age, cervical dilation, and maternal BMI as described in the previous text.
The logistic regression model coefficients were used to construct a risk score such that:
log(risk score) = 18.52 + 1.58 (HPD) + 1.62 (caput) – 0.57 (occiput posterior position) + 0.07 (maternal age) – 0.05 (maternal BMI) – 0.51 (gestational age) – 1.31 (prolonged labor) + 0.27 (cervical dilation)
Final risk score = exp(log[risk score]).
The risk score was applied to all patients in the validation set (n = 61) so that every patient had a risk score ( Table 2 ). The score was then dichotomized at its median (2.91) and a logistic regression model run to see whether continuous and dichotomized scores were significantly associated with vaginal delivery in the validation set ( Table 3 ). A receiver-operating characteristic (ROC) curve and corresponding area under the curve were generated using the pROC package in R to determine how well the continuous score predicts vaginal birth in the validation set ( Figure 3 ).
